Polly Clayden, HIV i-Base
Two cohort studies – from the Netherlands and France – showed a higher risk of IRIS with integrase inhibitor-based ART than with other regimens. Findings from both studies were presented at CROI 2017.
HIV treatment with integrase inhibitor (INSTI) based ART is recommended as preferred first-line in most high-income countries and is expected to become standard of care in many low- and middle-income countries over the next few years.
The use of INSTI based ART is associated with an accelerated viral load decline and faster CD4 increase compared with PI or NNRTI based ART.
IRIS is a pathological inflammatory response to antigens of opportunistic infections (OIs). People with CD4 counts of less 200 cells/mm3 when they start ART are at an increased risk for IRIS. Typically phase 3 registrational studies of new antiretrovirals exclude people with low CD4 counts and active OIs.
The Dutch findings were from a multicentre, retrospective observational study in the ATHENA cohort.  The investigators performed a chart review of all treatment-naive participants starting ART from 2009 onwards (when raltegravir was registered in the Netherlands) who were at risk for IRIS. This included people with: CD4 200 cells/mm3 or less at start of ART; on OI diagnosed before or after starting ART; and/or use of corticosteroids up to 12 months after starting ART; and/or died up to 12 months after starting ART.
The study used two definitions of IRIS.
- According French et al, 2004 (An atypical presentation of an OI or tumour in a patient responding to ART with decline in viral load or increase in CD4 count); and
- Diagnosed by the treating clinician (IRIS as most likely diagnosis in the patient file or IRIS in the differential diagnosis with initiation of treatment for IRIS – eg corticosteroids).
The primary endpoints were French IRIS and combined clinical or French IRIS in INSTI vs non-INSTI users (the investigators considered French IRIS to be more specific so participants with IRIS according to both definitions were categorised as French). Participants who switched from PI/NNRTI to or from INSTI were censored.
Of 18,355 participants in the ATHENA cohort, 369 met the study criteria and were included at time of analysis: 69 received INSTI based ART and 300 non-INSTI based. At HIV diagnosis, median CD4 count and viral load in the INSTI vs non-INSTI groups respectively was 36 vs 30 cells/mm3 and 446,694 vs 257,040 copies/mL. The majority of participants were men: respectively 74% and 83%. Median age overall was 43 years.
Incidence of French IRIS in the INSTI group was 19% and clinical IRIS 19%, compared to 9% and 7% in the non-INSTI group, with IRIS being three times more likely in the INSTI group according to either definition (OR: 3.25, 95%CI: 1.83 to 5.8).
Cox regression showed that INSTI use was independently associated with French as well as French/ clinical IRIS: HR 2.62 (95%CI: 1.35 to 5.1), p=0.0045, and HR 2.69 (95%CI: 1.63 to 4.44), p=0.0001, respectively.
IRIS was most frequently related to pneumocystis jirovecii pneumonia (PCP), candidiasis and mycobacterial infections.
Sex, age, ethnicity, mode of infection, calendar year, baseline CD4, CD4/CD8 ratio, highest viral load measurement, type of ART regimen, number and type of OI-events, time between start of OI treatment and start of ART and use of steroids for OIs, were not associated with an elevated HR for IRIS.
The investigators did not report any interaction between INSTI-use and any of the other characteristics that were significantly associated with IRIS.
The French study was from the Dat’AIDS cohort across French HIV centres that share the same electronic patient record system. 
The investigators selected participants starting ART with CD4 200 cells/mm3 or less (2010-2015) and admitted to hospital within six months and compared those receiving INSTI-based with non-INSTI regimens. Three HIV specialists blinded to the ART regimen performed the examination of the patients’ medical charts.
IRIS was defined as symptoms consistent with an infectious or inflammatory condition associated with a viral load drop of >2 log10 copies/mL not explained by a new infection, the clinical course of a previous infection, or side-effects, according to adapted ACTG IRIS criteria.
The study included 2287 participants: 398 received INSTI based ART and 1889 non-INSTI. Median age was 45 years and 63% were men. The third drug was a boosted PI in 65%, NNRTI in 12%, and INSTI in 12%. At ART initiation, the median CD4 count and viral load were 34 vs 84 cells/mm3, and 5.3 vs 5.2 log10 copies/mL and in the INSTI vs non-INSTI groups respectively.
Median viral load was lower in the INSTI group after three months of ART: 1.7 vs 2.1 log10 copies/mL, p<0.001.
IRIS occurred in 3% of participants in the INSTI group compared with 1.5% in the non-INSTI: OR 1.99 (95%CI: 1.09 to 3.47), p=0.04.
IRIS was most frequently related to tuberculosis, Mycobacterium avium and progressive multifocal leukoencephalopathy (PML).
As with the previous study co-factors such as age, sex and mode of infection were not associated with increased risk of IRIS.
These two reports are from cohorts and not randomised so come with all the associated potential for confounding with cohort studies.
The Dat’AIDS investigators noted that the relative infrequency of IRIS events in their study (1.8% overall) might be explained by the strict definition of IRIS they used and that they focused exclusively on severe IRIS needing hospitalisation. They also suggested it might be possible that because of known interactions between PIs and rifampicin that people with a pre-existing mycobacterial disease could be more likely to receive an INSTI based regimen. They did not demonstrate this in the study but probably lacked power to find any significant difference in the choice of first-line ART in IRIS cases.
Possible confounding aside, people starting INSTI based ART with a CD4 count of 200 cells/mm3 or less appear to be at increased risk for IRIS. This is important for low- and middle-income countries for which a dolutegravir based regimen is now recommended as an alternative first-line by WHO, and is predicted to become the preferred regimen as evidence gaps are filled.
Two ongoing large randomised trials of first-line dolutegravir vs efavirenz in as-close-as-possible-to-real-life unselected African patients that are likely to include people with low CD4 cell counts – ADVANCE (South Africa) and NAMSAL (Cameroon) – plan to have results in 2019 – but incidence of IRIS should be raised now as an ongoing concern for DSMBs for each study. 
The ADVANZ-4 Trial (Spain) is comparing first-line ART with dolutegravir and darunavir/ritonavir, both with abacavir and 3TC, in people with very low CD4 counts: below 100 cells/mm3 before treatment.  It is a small study with only 108 participants but is looking at immune reconstitution with results expected at the end of this year.
In the meantime, there needs to be careful monitoring of people with low CD4 counts at increased risk for IRIS in early dolutegravir adopter countries such as Botswana and Brazil.
- Wijting I et al. Integrase inhibitors are an independent risk factor for IRIS: an ATHENA cohort study. CROI 2017. February 13-16, 2017. Seattle. Poster abstract 731.
http://www.croiconference.org/sessions/integrase-inhibitors-are-independent-risk-factor-iris-athena-cohort-study (Abstract and poster link)
- Dutertre M et al. Initiation of art based on integrase inhibitors increases the risk of IRIS.
CROI 2017. February 13-16, 2017. Seattle. Poster abstract 732.
http://www.croiconference.org/sessions/initiation-art-based-integrase-inhibitors-increases-risk-iris (Abstract and poster)
- Clayden P. Fit for purpose: antiretroviral treatment optimisation. HIV i-Base. 12 February, 2017.
- US National Institutes of Health. Immune recovery in advanced, ARV-naive, HIV-1-infected individuals taking dolutegravir or ritonavir-boosted darunavir